yunwal

> economy is 100% automated?

People thought tractors would automate everything too. They didn’t. There’s more to life and work than being able to generate text.

> “AI model submits false tip”, it doesn’t say anything about a form

The article mentions the website name in the first sentence, and it’s trivial to see that you submit a tip by filling out a form.

> It's not our job to attempt to solve large-scale societal issues caused by technological development

Says who?

> Note that pretty much every time during the industrial revolution a technology replaced human effort, it ended up creating more jobs and wealth in the long term

Note that since the 1970s, and throughout the entire personal computing era, median wages have not outpaced inflation. Fewer Americans have access to good schools, median home buying age has increased, access to good healthcare has decreased. So I think there’s good reason to believe the “more jobs and wealth in the long term” trend has reversed for most people.

This is absolutely not true. The reason nearly every mathematician believes the Riemann hypothesis is true is because out of zillions of discovered non-trivial zeros, so far 100% have been located at (0.5, <prime>). If we had only discovered 3 such points, nobody would give it much credence.

> you can't run experiments on your own

Who says?

> there is a possibility that a clinical trial run by AI could hack the trial get something approved when it wasn't safe or effective

This possibility already exists and nothing about AI fundamentally changes it

> humans would follow the spirit of the trial process

P-hacking has been a problem well before LLMs were wearing diapers

I would say this statement is incorrect, and if I had to guess, it will be for the next 5 years. Google’s main product is ads, followed by search, gmail, and Google Docs suite, and Gemini is not on track to dethrone either any time soon.

I think eventually we’ll substitute code review with statistical evals for some cases. Similar to how we do drug trials.

I don’t think that’s “the whole shtick” anymore than it is for something like cars. There are valid ways to use it as a tool to provide net new value, and inevitably a whole bunch of people will end up using it as a lazy machine anyway, especially if their country/society benefits from people using it that way (Americans in both cases).

If it's a threat to people's livelihood or sense of meaning, it makes sense to consider how we can prevent massive numbers of people (especially incredibly smart people) from becoming disgruntled, depressed, or destitute. Typically societies with a lot of unhappy people do not function well (perhaps by definition), even if they have some technological edge.

I can’t imagine anyone thinking filling out a form requires scary levels of power. It actually makes the model look incredibly dumb. It’s probably a real, stupid thing that happened.

LLMs were also a nice example of this before 2018 or so, so I wouldn't necessarily count on quantum computing being a disappointment forever.

Right, they also seem to have hidden it from the Australian government a month after learning about it. That feels like it should qualify as conspiracy in any reasonable legal system

In my view, it's not that socialism with a lowercase s as defined by worker ownership of the means of production has gotten way more popular, it's that "Socialist" has shifted to mean New Deal Democrat as the Democratic party has shifted towards being the party of business interests.

If you look at what Mamdani, arguably the face of American Socialism, is doing in NYC, it's basically deciding to hold businesses accountable for breaking laws and noticing where markets have failed to provide core needs in the city.

> In my experience, the docs for something like setting up a dev env are typically greatly improved by the second person who sets up the dev env, not the personal who originally wrote the docs.

In my experience, this is also true of a lot of code as well. Your dev scripts should probably have much more relaxed standards than your service source or CI/CD. Ideally I could define merge requirements by directory without doing some weird shenanigans with the CODEOWNERS file and a bot.

Let's repeat the same circle. Why is customer service fundamentally a more controlled environment than anything else the author deems "outside of an LLMs capabilities"?

Nope, not having trouble following it, but the metaphor leaves out an important link between AI output and the AI itself. The question is, did the AI gain anything from this training? And the answer is, it did not, since the training used it's own answer with no other input. There's nothing in the water metaphor that draws this line.

Maybe a better metaphor is something involving drinking your own piss, but I don't feel the need to flesh that one out

I don't think anyone is saying it's poisonous here? The people were hired to do a job and they didn't do it. Feels like you're being disingenuous, if there's one thing i know about AI employees, it's that they use LLMs, like, all the time.

sure, but then why be bearish on LLMs? You don't even need to care about the code just check the boxes.

> those who need done a small set of narrowly defined tasks with existing clear guardrails: repetitive physical labor in a controlled environment, call center and customer service chat work, etc.

I have no idea how people can so confidently say that call center work is a “controlled environment” or “repetitive”. It’s almost by definition not repetitive or controlled. Customer support is what I go to when the controlled environment has failed

> If Jev is skipping generation entirely for a narrow structured task, of course it's faster

I think this is reasonable if people are actually using LLMs to solve this type of narrow structured task, which they are. The evidence is that every LLM provider has some method of forcing the output to conform to a json schema in their documentation.

> You mean they are Jewish

No you, the commenter said what they said. I honestly think this reply merits a ban or at least a warning from moderators

Literally no person on earth can beat a computer in chess. Practically no person on earth can really teach a computer anything about chess. For some crazy reason, people keep playing

> Take Software engineering for example, I can't believe there is a class of software engineers who wake up everyday and tell themselves, "today is the day I am going to automate the rest of my job".

And here I thought this was the whole point…

I think most people here get that this is the point?

The point beyond this one is that an AI proof doesn’t prevent humans from working on the problem, it destroys the current economic incentive to work on the problem. Perhaps we should rethink the current incentives. In order to make money as a chess player, you don’t need to beat AI, or ban AI from playing chess.

If mathematics took a similar approach (we don’t get paid for solving net-new problems, we get paid for enriching human understanding), then there’s no issue.

What I’m hearing is “it’s ok when other people lose their jobs…”

It’s not going to be ubiquitous? There hasn’t been a single frontier model where generation n costs less than generation n-1 to run. So the reasonable thing is to assume that GPT-7 will cost even more than GPT-6, and more and more of the frontier of knowledge will be locked behind a giant paywall. Participating in any field will mean ponying up to the oligarchs that own the infrastructure that runs the model.

> Why does the Vision Pro exist then?

They are thinking that VR headsets will not be niche products in the future

The apple watch was very niche when it came out as well, but their intention was absolutely to make a major product line

I would love it if I could use it as a second monitor when working in a coffee shop or library. Unfortunately universal control was pretty solid at launch and then just completely left to die